Numerical analysis on MUF process for flip chip packaging

Abstract

With merits like high I/O density, superior electrical and thermal performance, and small form factor, flip chip has become more and more widely used in electronic packages. Although the flip chip packaging process has been fairly improved, the conventional CUF (Capillary Underfill) process is a bottleneck that results in lower productivity and higher cost. MUF (Molded Underfill) process offers an alternative solution with advantages such as simplification of process, reduction of material cost and higher productivity. However, the extremely narrow gap between substrate and die makes it a big challenge to get optimal MUF material set and appropriate process conditions to cope with the serious void trapping issue. Besides, the process-induced warpage is also a problem that affects yield and reliability. In this paper, the MUF process of a flip chip package has been studied using numerical analysis method. The cure-kinetic, rheological and chemical shrinkage properties of MUF compound were measured by DSC (Differential Scanning Calorimeter), DMA (Dynamic Mechanical Analyzer) and universal testing machine, respectively. A global model and a sectional model were developed for MUF injection analysis to simulate the melt front advancement and estimate possible mold void distribution. Warpage analysis considering both chemical shrinkage and CTE (Coefficient of Thermal Expansion) mismatch was performed afterwards. The MUF injection analysis showed mold void generated near the center of underfill area. The warpage analysis indicated that stresses caused by chemical shrinkage and CTE mismatch acted in opposite directions, and the overall warpage was dominated by CTE mismatch. Simulation results of injection analysis were validated by short shot experiment and SAM (Scanning Acoustic Microscopy) test, respectively, and the warpage prediction was verified by measurement data. Finally, MUF process setting with four parameters was optimized based on the established numerical models, three indicators of mold quality had been improved by 9.10%, 8.95% and 20.6% respectively.